bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Open Access

RESEARCH ARTICLE Palaeobiological inferences based on long bone epiphyseal and diaphyseal structure - the forelimb of xenarthrans (Mammalia)

Eli Amson & John A Nyakatura

Cite as: Amson E and Nyakatura JA. (2018). Palaeobiological inferences based on long bone epiphyseal and diaphyseal structure - the forelimb of xenarthrans (Mammalia). bioRxiv, 318121, ver. 5 peer-reviewed and recommended by PCI Paleo. DOI: 10.1101/318121

Peer-reviewed and recommended by Peer Community in Paleontology

Recommendation DOI: 10.24072/pci.paleo.100001 Published: 21 September 2018 Recommended by: Alexandra Houssaye Based on reviews by: Andrew Pitsillides and an anonymous reviewer

PEER COMMUNITY IN PALEONTOLOGY bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

Preprint uploaded to Biorxiv on 21 September 2018 Version 5, © 2018 Amson, Nyakatura CC-BY-NC-ND 4.0

Palaeobiological inferences based on long bone epiphyseal and diaphyseal structure - the forelimb of xenarthrans (Mammalia)

Eli Amson1,2, John A Nyakatura2

1Museum für Naturkunde, Leibniz-Institut für Evolutions- und Biodiversitätsforschung, Berlin, Germany; [email protected] 2AG Morphologie und Formengeschichte, Institut für Biologie & Bild Wissen Gestaltung. Ein interdisziplinäres Labor, Humboldt-Universität, Berlin, Germany; [email protected]

ORCID 0000-0003-1474-9613 (Eli Amson) 0000-0001-8088-8684 (John A Nyakatura)

This article has been peer-reviewed and recommended by Peer Community in Paleontology (https://dx.doi.org/10.24072/pci.paleo.100001).

ABSTRACT Trabecular architecture (i.e., the main orientation of the bone trabeculae, their number, mean thickness, spacing, etc.) has been shown experimentally to adapt with great accuracy and sensitivity to the loadings applied to the bone during life. However, the potential of trabecular parameters used as a proxy for the mechanical environment of an organism’s organ to help reconstruct the lifestyle of extinct taxa has only recently started to be exploited. Furthermore, these parameters are rarely combined to the long-used mid-diaphyseal parameters to inform such reconstructions. Here we investigate xenarthrans, for which functional and ecological reconstructions of extinct forms are particularly important in order to improve our macroevolutionary understanding of their main constitutive clades, i.e., the Tardigrada (sloths), Vermilingua (anteaters), and Cingulata (armadillos and extinct close relatives). The lifestyles of modern xenarthrans can be classified as fully terrestrial and highly fossorial (armadillos), arboreal (partly to fully) and hook-and-pull digging (anteaters), or suspensory (fully arboreal) and non-fossorial (sloths). The degree of arboreality and fossoriality of some extinct forms, “ground sloths” in particular, is highly debated. We used high-resolution computed tomography to compare the epiphyseal 3D architecture and mid-diaphyseal structure of the forelimb bones of extant and extinct xenarthrans. The comparative approach employed aims at inferring the most probable lifestyle of extinct taxa, using phylogenetically informed discriminant analyses. Several challenges preventing the attribution of one of the extant xenarthran lifestyles to the sampled extinct sloths were identified. Differing from that of the larger “ground sloths”, the bone structure of the small-sized Hapalops (Miocene of Argentina), however, was found as significantly more similar to that of extant sloths, even when accounting for the phylogenetic signal. Keywords: Bone structure; Forelimb; Locomotion; Palaeobiological inferences; Trabeculae; Xenarthra

(Ruff et al. 2006). This was argued for trabecular INTRODUCTION bone, which reacts to loading with great accuracy and sensitivity (Barak et al. 2011). This was also argued Bone structure is intensively studied in analyses for cortical bone, even though the latter is expected concerned with functional anatomy because it is to be less plastic, at least in part due to its lower argued to be extremely plastic. While a genetic remodeling rate (see review of Kivell, 2016). blueprint influences bone structure, it has been Comparative studies focusing on either trabeculae or shown to adapt during life (and especially at an early cortical structure intend to leverage this great ontogenetic stage) to its mechanical environment plasticity to associate structural phenotypes to

1 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

lifestyles or functional uses of a limb. This has been however, that Tsegai et al. (2017), also used this achieved in some analyses (as recently exemplified holistic analysis and performed a Principal by Georgiou et al. 2018; Ryan et al. 2018; Tsegai et Component Analysis (even though in that case the al. 2018) but not all of them (see review of Kivell focus was on trabecular bone architecture and 2016), suggesting that some confounding factors are cortical bone thickness at the articular surface). likely to be at play, and more generally that the Skinner et al. (2015) and Stephens et al. (2016) also approach is limited. For trabecular bone in particular, used Gross et al. (2014)’s method, but focused on important intraspecific variation has been trabecular architecture only. This approach is documented (e.g., in Pongo; Tsegai et al. 2013; particularly relevant for medium- to large-sized Georgiou et al. 2018). Nevertheless, the fact that mammals such as Pan or Homo, for which the some analyses successfully distinguished ecological epiphyses include a complex trabecular architecture groups might indicate that broad differences of bone with distinct zones of different arrangement (such as structure among lifestyles can exceed, at least in the so-called vertical and horizontal trabecular some cases, individual variability. Because fossil columns in the femoral neck; Hammer 2010). One bone cross-sections at mid-diaphysis have been can note that an entirely different approach, not produced for over a century and a half (Kolb et al. relying on the measurement of these parameters, but 2015), a large number of mid-diaphyseal data related on micro-finite element analysis, was also applied to to extinct taxa have been acquired, and successfully a primate (Huynh Nguyen et al. 2014). To our exploited for palaeobiological inferences (e.g., knowledge, epiphyseal trabecular and mid- Germain & Laurin, 2005). Fossil three-dimensional diaphyseal parameters have never been combined in (3D) trabecular architecture has been much less a functional analysis about non-primate taxa, and no investigated, as, to our knowledge, only few studies analysis used both trabecular and cross-sectional have been published, which are all focussing on parameters in the same discriminant test. primates (DeSilva & Devlin 2012; Barak et al. 2013; References to bone structure in “ground sloths”, Su et al. 2013; Skinner et al. 2015; Su & Carlson Megatherium in particular, date back to the 19th 2017; Ryan et al. 2018). century (Owen 1861). But it is only fairly recently that In general terms, it is assumed that the diaphysis quantification of bone structure was performed of long bones tends to be exposed to mostly bending (Straehl et al. 2013; see review of Amson & and torsion, and to a lesser extent axial compression Nyakatura 2017). Straehl et al. (2013) examined (Carter & Beaupré 2001). On the other hand, the compactness profile of a mid-diaphyseal section in architecture of epiphyseal trabeculae is usually the limb long bones of various extant and extinct related to compressive and tensile strains (Biewener xenarthrans. They found that most armadillos were et al. 1996; Pontzer et al. 2006; Barak et al. 2011). characterized by a humeral mid-diaphysis that is Trabecular and cortical compartments are hence relatively more compact than that of the femur. expected to have distinct mechanical properties, Subsequently, Amson et al. (2017a) studied the which do not necessarily co-vary. To combine them epiphyseal trabecular architecture in extant in a single analysis, it can therefore be argued that xenarthrans, and found that some parameters, the the structural parameters deriving from these two degree of anisotropy (DA) in particular, differed types of structures should be considered as distinct among functional categories. (univariate) variables. Because trabecular and Indeed, xenarthrans are marked by distinct cortical structures have independently yielded a lifestyles that can be used to define functional functional signal, such a combined analysis could categories. Extant xenarthrans were categorized by potentially help in our endeavours to associate a bone Amson et al. (2017a) as fully arboreal and non- overall structure to a loading regime, and, eventually, fossorial (extant sloths), intermediate in both a function. This combined analysis has previously fossoriality and arboreality (anteaters), or fully been achieved, on extant taxa, via different terrestrial and fossorial (armadillos), and several approaches. Based on epiphyseal regions of interest fossorial classes were recognized among the latter. (ROIs) and mid-diaphyseal sections, Shaw & Ryan Partly following their expectations, Amson et al. (2012) examined both compartments in the humerus (2017a) recovered that the armadillos (and in and femur of anthropoids (see also Lazenby et al. particular the more highly fossorial ones) differ in their (2008) for handedness within humans). They greater DA for instance. The latter can be expected measured individual trabecular and mid-diaphyseal to be associated with the presence of one main parameters, but did not combine the latter in a single loading direction in these highly fossorial taxa (as test. Another approach, termed ‘holistic analysis’ opposed to various equally marked directions in taxa (Gross et al. 2014), was used in Pan and Homo whole for which the forelimb is arguably facing a less bones or epiphyses, but parameters were not used stereotypical main loading). Similarly, one could conjointly to discriminate functional groups in the expect those taxa of which the long bone in question statistical assessment either. It is noteworthy,

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 2 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

experiences one main bending direction to be Delsuc et al. 2001), and the two extant genera, characterized by a more elliptical cross-sectional Bradypus (three-toed sloths) and Choloepus (two- shape at mid-diaphysis (CSS, see below), with the toed sloths), most likely acquired their highly derived section’s major axis aligned along that direction (as lifestyle convergently (Nyakatura 2012; Coutier et al. the major axis indicates the direction of the greatest 2017). Most aspects of the biology of “ground sloths” bending rigidity; Ruff & Hayes 1983). Because no exceed in disparity those of their extant kin. They significant differences were recovered in the mid- were found from Alaska (Stock 1942) to diaphyseal global compactness between fossorial southernmost South America (and potentially and non-fossorial talpid moles (Meier et al. 2013), it Antarctica; Carlini et al. 1990; Gelfo et al. 2015) seems that a simple relation between this parameter Various feeding habits, such as bulk-feeding or and a loading scheme associated with fossorial selective-feeding are purported (Bargo & Vizcaíno activity should not be expected (see also Straehl et 2008). The lifestyle of most extinct sloths is al. 2013). reconstructed as terrestrial (but see Thalassocnus for an (semi-)aquatic lineage; Amson et al. 2015b). For extant xenarthrans, the functional categories Furthermore, some “ground sloths” contrast with their mentioned above mostly match the phylogeny, i.e., extant relatives in reaching large body sizes (up to most categories are aggregated into clades. several tones; Fariña et al. 1998). However, this is likely not true if one includes the extinct xenarthrans, the “ground sloths” in particular, The fossil record of early (Palaeocene-Eocene) because their lifestyle was interpreted as different xenarthrans and especially that of sloths, is rather from that of their closest relatives, the “tree sloths”. poor (Gaudin & Croft 2015). It is therefore hard to Lifestyle reconstruction of extinct xenarthrans dates reconstruct the ancestral lifestyle of Tardigrada, and back to the 18th century (see review of Amson & more generally Xenarthra. To date, no extinct sloths Nyakatura 2017). Various methods were employed to have been reconstructed to have had a suspensory infer the lifestyle of extinct xenarthrans. So far, they posture and locomotion resembling their extant kin all relied on bone (and tooth) gross morphology, (Pujos et al. 2012). But, because their gross anatomy involving approaches such as comparative functional was considered as similar to that of extant anteaters, morphology (Amson et al. 2015a), biomechanical Matthew (1912) argued that Hapalops, for instance, modelling (Fariña & Blanco 1996) or muscle was partly arboreal. Such a lifestyle was of course not reconstruction (Toledo et al. 2013). This was found to considered for larger taxa (but see translation of Lund be challenging, partly because of the lack of modern in Owen (1839) for an early opposite view). However, analogues for some taxa (Vizcaíno et al. 2017), and digging capabilities, as well as bipedal stance and/or partly because of the autapomorphic nature of locomotion, was proposed for several medium-sized several of the xenarthran traits. This, along with the (e.g., Glossotherium) to giant-sized (e.g., fact that functional categories mostly match Megatherium) “ground sloths” (Bargo et al. 2000; phylogeny, makes disentangling the phylogenetic Patiño & Fariña 2017). For the present analysis, we and functional signals difficult (Amson et al. 2017a). were able to sample small-sized as well as large- Bone structure was argued to be extremely plastic sized “ground sloths.” The estimated body sizes of and found in xenarthrans in particular to be mostly the latter exceed that of extant xenarthrans by two devoid of phylogenetic signal (and when a significant orders of magnitude (see below for body mass signal is found, it is likely due to the matching estimates). Because this has already been pointed between functional categories and clades; Amson et out as a challenge for the reconstruction of extinct al. 2017a). The ecophenotypic nature of bone xenarthrans’ lifestyles (Vizcaíno et al. 2017), and structure traits (which are defined as because size might be correlated to at least some "biomechanically informative phenotypically plastic"; bone structure parameters, potential challenges Ryan et al. 2018) is the rationale behind the present inherent to the taxa and parameters we studied will endeavour. be discussed. The aim of this study is to quantify bone diaphyseal and trabecular structure in “ground sloths” in order to infer their lifestyle. Given the disparate gross morphology of xenarthrans (e.g., the humerus is extremely slender in extant sloths and particularly MATERIAL AND METHODS stout in most armadillos, see Mielke et al. 2018a), we believe that studying easily comparable and arguably Specimen and scanning procedure ecophenotypic traits such as bone structure parameters is highly relevant for this purpose. Extant The dataset of Amson et al. (2017a), which sloths represent but a remnant of the overall diversity consists of extant skeletally mature wild-caught of Tardigrada (also termed Folivora or Phyllophaga; xenarthrans, was extended by several extinct sloths roughly spanning the whole body size range of the

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 3 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

m a halu p m a l tu

CingulataCingulata y s s toce y p lus s atus

VermilinguaVermilingua y us g nctus i a tridactyl d c g ri i b maximu

“ground“ground sloths” y s h hractus villosus hractus phorus truncatus

Tardigrada: p

“tree“tree ssloths”loths” y pus torquatus otherium robu y todon armatu ss ypeutes tr clopes didact l s haetophractus vellerosus haetophractus retusus alyptophractus y uphractus sexcinctus uphractus riodonte o amandua tetradact e Dasypus novemcinctus Dasypus hybridus sexcinctus Euphractus villosus Chaetophractus Zaedyus pichiy vellerosus Chaetophractus Chlamyphorus truncatus retusus Calyptophractus Priodontes maximus tricinctus Tolypeutes Cabassous tatouay tetradactyla Tamandua Myrmecophaga tridactyla Cyclopes didactylus Choloepus didactylus torquatus Bradypus variegatus Bradypus Dasypus novemcinctus Dasypus E Chaeto Zaedyus pichiy C Chlam C P T Cabassous tatoua T Myrmecopha C Choloepus didactylus Brad varie Bradypus Scelidotherium leptocephalum Glossotherium robustum Lestodon armatus 0 Scelidotherium le Glo L

Piacenzian Pliocene Zanclean Messinian

Tortonian

10 s p

Serravallian lo a Miocene

Langhian ap Neogene Hapalops H Burdigalian 20 Aquitanian

Chattian

Oligocene 30 Rupelian

Priabonian

40 Bartonian

Eocene Lutetian Paleogene

50 Ypresian

Thanetian 60 Paleocene Selandian

Danian

Upper Maastrichtian 70 retaceous

C Campanian Figure 1. Timetree depicting the time-calibrated phylogenetic relationships of the xenarthrans included in the phylogenetically flexible linear discriminant analyses. See Material and Methods section for the sources used to build the timetree. group: the small-sized (ca. 38 kg; Bargo et al. 2012) specimens are skeletally mature (a few specimens Hapalops sp. (Santa Cruz Formation, Early Miocene, showed a remnant of epiphyseal line, see below) and ca. 17 Ma; Perkins et al. 2012), the medium-sized (ca. did not present apparent bone diseases (which were 200 kg; Smith et al. 2003) Valgipes bucklandi (Lagoa also criteria of selection for the extant species, see Santa, Brazil, Pleistocene; the sampled specimen Amson et al. 2017a). All fossils were scanned (micro MNHN.F.BRD29 is labelled Ocnopus gracilis, which computed tomography, µCT) using a v|tome|x 240 L is now viewed as a junior synonym; Cartelle et al. system (GE Sensing & Inspection Technologies 2009), Scelidotherium leptocephalum (ca. 1000 kg; Phoenix X|ray) at the AST-RX platform of the Vizcaíno et al. 2006) and Glossotherium robustum Museum national d’Histoire naturelle (Paris, France). [ca. 1200 kg (Vizcaíno et al. 2006); both from According to the methodology and results of Amson ‘Pampean’, Argentina and Tarija, Bolivia, both et al. (2017a), we focused our data acquisition of the Pleistocene], as well as the large-sized Lestodon trabecular parameters on the humeral head and armatus [ca. 3200 kg (Vizcaíno et al. 2006); radial trochlea regions of interest (ROIs; see below). ‘Pampean’, Argentina, Pleistocene] and Megatherium Mid-diaphyseal parameters were acquired for these americanum [ca. 4000 kg (Fariña et al. 1998); two bones and for the third metacarpal (Mc III) in all ‘Pampean’, Argentina, Pleistocene]. The sampled species, when available. See Table 1 for the list of

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 4 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

Table 1. List of fossils with type of data acquired for each bone.

Species Specimen number Data type

Humerus Radius Mc III

Hapalops sp. MNHN.F.SCZ166 - 72%MD; 100%TA -

Hapalops sp. MNHN.F.SCZ164 35%MD; 72%TA; (39%TA) - -

Hapalops sp. MNHN.F.SCZ162 50%MD; 35%MD; (39%TA) - -

Lestodon armatus MNHN.F.PAM754 - 50%MD

Lestodon armatus MNHN.F.PAM755 - - 50%MD

Lestodon armatus MNHN.F.PAM95 100%TA - -

Glossotherium robustum MNHN.F.PAM756 QO 50%MD; 100%TA -

Glossotherium robustum MNHN.F.PAM141 - - 50%MD

Scelidotherium leptocephalum MNHN.F.PAM236 50%MD - -

Valgipes bucklandi MNHN.F.BRD29 - - 50%MD

Megatherium americanum MNHN.F.PAM753 - - 50%MD

Megatherium americanum MNHN.F.PAM758 - QO -

Footnotes. Abbreviations: 'n'-MD, mid-diaphyseal data, with 'n' the position of the sampled cross-section expressed as the length percentage from the proximal end; 'n'-TA, trabecular architecture data, with 'n' the cropping coefficient that was used, if any (see Material and Methods section); QO, only qualitative observations were performed. skeletal elements sampled for each extinct species, intermediate anteaters, and fully terrestrial and along with ROIs for which data were successfully fossorial armadillos. acquired [see also Amson et al. (2017a), for sample size and scanning procedure of the extant species specimens]. For the included specimens, scanning resolution ranged from 0.03 to 0.123 mm (depending Qualitative observation of the diaphyseal on the size of the specimens). Relative resolution, used to assess if the employed resolution is adequate structure to analyse trabecular bone (mean trabecular Raw image stacks were visualized with the Fiji thickness divided by resolution) ranged from 5.1 to package (ImageJ2 v. 1.51n and plugins; Schindelin et 11.5 pixels/trabecula. This is considered as al. 2012, 2015; Schneider et al. 2012). The appropriate (Sode et al. 2008; Kivell et al. 2011; ‘Orthogonal Views’ routine was used to compute Mielke et al. 2018b). Scanning resolution (and longitudinal sections. Sedimentary matrix prevented relative resolution for the trabecular ROIs) for each satisfying segmentation for some specimens but at specimen can be found in Supplementary Online least some qualitative observations were possible for Material (SOM) 1. For this first endeavour of all specimens (see Table 1). palaeobiological reconstruction of “ground sloths” lifestyle based on bone diaphyseal and trabecular Trabecular parameters structure, we compared the parameters yielded by the fossils to those of the extant specimens, using the We followed the methodology of Amson et al. same lifestyle categories as defined by Amson et al. (2017a), which involves the use of the BoneJ plugin (2017a), i.e., the fully arboreal extant sloths, (Doube et al. 2010) for Fiji. In brief, bones were first placed in the same standard orientation. Then, ROIs

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 5 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

were selected in the centre of the studied epiphyses, with the ‘Fit Sphere’ routine of BoneJ (see Amson et al. 2017a: fig. 2 and Additional files 3, 4). ROI were selected to be as large as possible but without including cortical bone. We used the ‘Orthogonal Views’ routine of Fiji to ascertain that the centre of the ROI was precisely located at the centre to the studied epiphysis along the mediolateral, anteroposterior, and proximodistal directions. The resulting substack was then thresholded (‘Optimise Threshold> Threshold Only’ routine) and purified (‘Purify’ routine). Finally, trabecular parameters were measured. Given the results of Amson et al. (2017a), we focused on the degree of anisotropy (DA), main direction of the trabeculae (MDT), bone volume fraction, BV/TV, connectivity density (Conn.D), trabecular mean thickness (Tb.Th), trabecular mean spacing (Tb.Sp), bone surface area (BS). Other trabecular parameters routinely acquired, however, can also be found in SOM 1. For some specimens, the lack of contrast between bone and the sedimentary matrix prevented accurate bone segmentation (see Table 1). Thresholding (see above) was successfully performed for the rest of the specimens; some of the latter, however, required manual removal of a few sedimentary particles (using the un-thresholded stack to recognize them). The humerus of two specimens of Hapalops showed a slight remnant of epiphyseal line. A smaller ROI was hence defined to exclude this line (which would have biased the measurements) by cropping isometrically (in 3D) the substack (custom ImageJ script, SOM 2). The cropping coefficient (MNHN.F.SCZ162: 39%; MNHN.F.SCZ164: 72%) was then applied to the whole dataset and trabecular parameters were acquired anew. The means of the latter were compared to the initial parameters. For the dataset cropped at 72%, differences were found as Figure 2. Qualitative observations of diaphyseal minor (similar MDT; DDA = 3%; DBV/TV < 1%; structure in xenarthrans. Longitudinal sections of humeri DConnD <1%), while for the dataset cropped at 39%, (A-C, E-F, all from CT-scans), tibia (D, ‘natural’ section), differences were more important (MDT of opposing and radius (G, from CT-scan). A, Chaetophractus direction; DDA, 13%; DBV/TV = 2%; DConnD, 4%). vellerosus (ZSM 1926-24); B, Priodontes maximus (ZSM Because it was exceeding a difference of 5% for at 1931-293); C, Myrmecophaga tridactyla least one parameter value, we did not analyse further (ZMB_MAM_77025); D, Nothrotherium maquinense the latter dataset (and excluded MNHN.F.SCZ162 (MCL 2821); E, Choloepus didactylus from the analysis of trabecular architecture). A (ZMB_MAM_35825); F, Glossotherium robustum (MNHN.F.TAR 767); G, Lestodon armatus remnant of epiphyseal line was also observed in (MNHN.F.PAM 754). Scale bars: A-E, 1 cm; F-G, 10 cm. Glossotherium robustum MNHN.F.PAM756, but in its case only qualitative observations were made. as the midpoint between most proximal and most distal points of either articular surfaces. Several sampled fossils did not preserve the mid-diaphysis. To compare them to the rest of the specimens, the Mid-diaphyseal parameters latter were re-sampled at the level closest to mid- The same standardly oriented µCT-scan stacks diaphysis preserved by each of those fossils (as (see above) were used for the acquisition of mid- assessed by superimposition with a complete diaphyseal parameters. Using Fiji, a cross-section specimen of the same species; MNHN.F.CSZ164 was selected at mid-diaphysis; the latter was defined (humerus): 35% from proximal end;

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 6 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

MNHN.F.CSZ166 (radius): 72% from the proximal Once the diaphyseal cross-sections were end; see Table 1). selected, they were thresholded automatically (see above), but we manually checked the resulting image, which, in a few instances, required a manual Table 2. Mean values of diaphyseal parameters of interest correction of the levels. The whole sectional area for each lifestyle category and extinct taxon. (WArea), global compactness (GC; both acquired CSA GC CSS with a custom ImageJ script, SOM 3), and cross- (mm2) (NU) (NU) sectional parameters of the ‘Slice Geometry’ routine Mc III diaphysis, of BoneJ (Doube et al. 2010) were acquired. For the following analyses, we focused on cross-sectional 50%MD area (CSA) and the ratio of second moment of area Armadillos 26.2 69.2 2.0 around major to minor axes (Imax/Imin), also termed Anteaters 56.6 73.8 2.1 cross-sectional shape (CSS). If the ratio is close to one, CSS will usually be roughly circular. Values Extant sloths 17.6 62.7 1.9 above one will entail increasingly elliptical shapes. Lestodon 1239.6 64.1 1.9 The other diaphyseal parameters, however, can also Glossotherium 1017.8 78.1 2.9 be found in SOM 1. Because, if normalized with WArea (see below), CSA would be redundant with Megatherium 2100.8 74.8 1.3 GC, it will only be used as a potential body size proxy. Valgipes 590.9 69.3 3.8

Humeral diaphysis, 50% Statistics Armadillos 65.8 68.5 4.4 The statistical analysis was performed using R Anteaters 144.8 66.5 3.0 version 3.4.3. Amson et al. (2017a) accounted for Extant sloths 59.2 72.8 1.2 size effects by computing a phylogenetically informed Hapalops 229.3 89.8 1.2 linear regression for each parameter, against a size proxy. If the regression was found as significant, its Scelidotherium 2780.8 80.9 2.6 residuals were used as the ‘size-corrected’ Humeral diaphysis, parameter. But the size of “ground sloths”, well 35% exceeding for most of them that of extant Armadillos 44.2 46.5 2.6 xenarthrans, could bias such a procedure. Indeed, the slightest error on the regression coefficients Anteaters 117.6 57.5 1.8 estimation would likely involve drastically different Extant sloths 63.0 64.8 1.2 residuals for those outlying taxa (see also Hapalops 235.3 75.3 1.8 Discussion). We therefore favoured, for the present Radial diaphysis, analysis, to normalize those parameters that have a 50% dimension by dividing the trait value by a body size proxy (raised to the same dimension). As body size Armadillos 17.2 89.0 2.7 proxies, we considered the specimen-specific TV (for Anteaters 58.0 83.1 2.2 trabecular parameters) and WArea (mid-diaphyseal parameters) or body mass (BM; species averages, Extant sloths 31.7 77.2 2.2 because unknown for most collection specimens). Lestodon 1474.7 71.3 5.0 Species body masses were taken from the AnAge Glossotherium 788.3 67.7 4.0 database (Tacutu et al. 2013) and additional sources Radial diaphysis, when necessary (Vizcaíno et al. 1999; Hayssen 72% 2010; Abba & Superina 2016; Smith & Owen 2017) for the extant species and from the specific sources Armadillos 28.4 76.2 3.8 mentioned above for the extinct taxa. The coefficient Anteaters 69.1 77.2 2.9 of determination of regressions against a parameter well known to correlate with size (Tb.Th for trabecular Extant sloths 35.1 71.3 5.1 parameters and CSA for mid-diaphyseal parameters) Hapalops 92.9 79.8 6.3 indicated that BM was more representative of the sample variance for the trabecular parameters, while Footnotes. Percentage indicates the position of the it was WArea in the case of mid-diaphyseal sampled cross-section, expressed as the length parameters. They were accordingly used as body percentage from the proximal end. Abbreviations: NU, no size proxies in the subsequent analyses. units.

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 7 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

Besides univariate comparisons, we performed we were not able to acquire all parameters for each linear discriminant analyses to infer the most likely of them (depending on the successfully processed lifestyle of extinct species. Both trabecular and mid- skeletal elements and ROIs, see Table 1). To diaphyseal parameters of the humerus and radius phylogenetically inform these analyses, we used the were conjointly used in these analyses (parameters function pFDA (Motani & Schmitz 2011; latest version from the Mc III were not included because of their lack available on github.com/lschmitz/phylo.fda). This of discrimination power, see Results). To account for ‘phylogenetic flexible discriminant analysis’ uses the the great body size disparity of the studied taxa, it is optimised value of Pagel’s Lambda to account for the the ‘size-normalized’ parameters that were used (raw phylogenetic signal (Pagel 1999). As implemented value divided by the relevant body size proxy if here, the latter can span from 0 to 1, respectively parameter not dimensionless, see above). One denoting absence of phylogenetic signal and trait analysis per extinct taxon was performed, because evolution consistent with a Brownian motion model of

A B Figure 3. Univariate n=21 n=9 n=4 n=1 n=1 n=1 n=1 n=21 n=9 n=4 n=1 n=1 n=1 n=1 comparisons of mid- diaphyseal parameters. A, Mc III Global Compactness (GC); B, Mc III cross- sectional shape (CSS); C, humeral GC; D, humeral CSS; E, radial GC; F, radial M3MS GC CSS. Thresholded mid- M3MS CSS diaphyseal virtual sections are depicted for the extinct sloths. Abbreviations: ant,

anteaters; arma, armadillos; 1.5 2.0 2.5 3.0 3.5 4.0 4.5

Glos, Glossotherium; Hapa, 40 50 60 70 80 90 Hapalops; Lest, Lestodon; arma ant sloth Val Glos Lest Meg arma ant sloth Val Glos Lest Meg Meg, Megatherium; Sce, Scelidotherium; sloth, extant C D sloths. n=20 n=9 n=5 n=1 n=1 n=20 n=9 n=5 n=1 n=1 HMS GC (raw) HMS CSS (raw) 60 65 70 75 80 85 90 123456

arma ant sloth Hapa Sce arma ant sloth Hapa Sce E F n=23 n=8 n=4 n=1 n=1 n=23 n=8 n=4 n=1 n=1 RMS GC (raw) RMS CSS (raw) 234567 70 75 80 85 90 95

arma ant sloth Lest Glos arma ant sloth Lest Glos

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 8 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

evolution. The rest of the pFDA works as a ‘traditional’ RESULTS discriminant analysis. The training data, stemming in our case from the extant xenarthrans, were classified according to the three main lifestyles, i.e., ‘armadillo’, Qualitative observations of diaphyseal ‘anteater’, and ‘extant sloth’. The test data relates to structure the sampled extinct sloths. If not already normally In the humerus of small armadillos and anteaters, distributed (as indicated by a Shapiro test), the the medullary cavity is mostly devoid of spongy bone parameters were log-transformed (and Shapiro tests (with just a few isolated trabeculae, e.g., were run again to confirm normality). Collinear Chaetophractus vellerosus ZSM-1926-24, Fig. 2A; variables (highly correlated variables as indicated by Cyclopes didactylus, ZMB_MAM_3913). In larger a correlation above 0.9) were excluded. members of these clades, the medullary cavity is filled The timetree used to phylogenetically inform the throughout the proximodistal length of the diaphysis tests was based on that used by Amson et al (2017a) by a more or less dense spongiosa (e.g., Priodontes (which is based on Gibb et al. 2016), and was maximus ZSM-1931-293; Myrmecophaga tridactyla, completed with the extinct taxa. The relationships ZMB_MAM_102642; Fig. 2B-C). In extant sloths, a between the main clades follow Amson et al. (2017b). spongiosa can be observed in most of the diaphysis The split between Mylodontidae (represented by (Bradypus; n=4) or throughout its length (Choloepus, Lestodon) and the other Eutardigrada (all sloths but Fig. 2E; n=4), but a central region free of trabeculae Bradypus) was set according to the age of the oldest subsists. The medullary cavity of the whole diaphysis fossil pertaining to the clade (Octodontotherium, ca. is full of spongy bone in Glossotherium (n=1; Fig. 2F). 29 Ma; Flynn & Swisher 1995; Kay et al., 1998) and It is nearly full in Scelidotherium, with just a small is thus conservative (Fig. 1). But one can note that central free region subsisting (n=1). For Hapalops, a this age is roughly as old or older that the recent clear assessment cannot be given due to the molecular estimations of the divergence time preservation of the specimens at hand between the two genera of extant sloths (Slater et al. (MNHN.F.SCZ162 seems to show a free medullary 2016; Delsuc et al. 2018). The age of divergence cavity, but MNHN.F.SCZ164, which only preserves between Lestodon and Glossotherium was set the proximal third of bone, shows a medullary cavity according to the age of Thinobadistes (Hemphillian, full of spongy bone). The whole diaphysis of the larger ca. 9 Ma; Woodburne 2010), which is more closely sloths Megatherium and Lestodon were not related to Lestodon than Glossotherium according to observed, but it is noteworthy that their epiphyses are Gaudin (2004). Extinct sloths were placed according filled with dense spongiosa (each n=1). to their known geological ages (see above; for The radius of extant xenarthrans shows the same Pleistocene taxa, a relatively young age of 0.1 Ma pattern as the humerus. In Glossotherium, Lestodon, was arbitrarily given. Length of the branches leading and Megatherium, the medullary cavity of the whole to nodes of unknown ages, which are in direct relation radial diaphysis is essentially full of spongy bone (Fig. to extinct taxa, and from these to terminal extinct taxa, 2G; no data for Hapalops for which the entire radial were arbitrarily set to 1 and 0.1 Ma, respectively. epiphysis could not have been sampled). Caution should be taken regarding the phylogenetic scheme used herein, because recent developments (yet to be published) in phylogenetic analyses of xenarthrans, which involve ancient DNA, might imply Univariate comparisons significant alterations of our understanding of sloths’ systematics (R.D.E. MacPhee, pers. comm., 2018). The structure of the Mc III of extant species did not differ notably among the lifestyle categories (Fig. 3A- B; Table 2). There is only a tendency for the anteaters Institutional abbreviations and armadillos to have a more compact mid- diaphysis (Fig. 3A). Mc III structure was therefore not MCL, Museu de Ciencias Naturais da Pontifícia further studied, and not included in the discriminant Universidade Católica de Minas Gerais, Belo analyses (see below). One can note, however, that Horizonte, Brazil; MNHN.F, Muséum national some armadillos have an outlyingly high CSS (i.e., d’Histoire naturelle, Paris, France, Palaeontology very elliptic cross-section) at mid-diaphysis (Fig. 3B; collection; ZMB_MAM, Museum für Naturkunde the single most elliptic value is found in the Berlin (Germany), Mammals Collection; ZSM; subterranean Calyptophractus retusus ZSM-1961- Zoologische Staatssammlung München, Germany. 316). A great disparity of CSS at this location is found in extinct sloths, with the value of Valgipes falling among the outlying armadillos just mentioned, and that of Megatherium being the single lowest (i.e., most circular cross-section).

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 9 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

The humeral diaphysis in Hapalops is remarkably sampled at 72% of diaphyseal length) falls among the compact. At mid-diaphysis (n=1), it features the distribution of armadillos, being slightly higher than highest GC value of the whole dataset (Fig. 3C; Table extant sloths’ values (Table 2). The GC of 2). At 35% of the diaphyseal length (from the proximal Glossotherium and Lestodon at radial mid-diaphysis end, level which was sampled to include fragmentary is very low, which agrees with the tendency observed fossils, see Material and Methods; n=2), Hapalops in extant sloths (Fig. 3E). The CSS at that location is falls in the uppermost distribution of the extant sloths, found as rather homogenously low among extant which does not markedly differ from that of armadillos xenarthrans, except for two armadillos with outlying or anteaters. The CSS at humeral mid-diaphysis high values. Glossotherium and Lestodon fall beyond distinguishes quite clearly the functional categories, the distribution of most extant xenarthrans, their CSS with high values (i.e., elliptical cross-sections) in being only tied or exceeded by the two outlying armadillos, intermediate values in anteaters, and low armadillos (Fig. 3F). values (i.e., round cross-sections) in extant sloths. In Regarding the trabecular architecture parameters, Hapalops, this parameter falls among the particularly only the degree of anisotropy (DA) will be presented tight range of extant sloths (Fig. 3D), but among that with univariate comparisons, as it was singled out as of anteaters at 35% of the diaphyseal length. In the most functionally informative of these parameters Scelidotherium (n=1), the GC of the humerus at mid- in extant xenarthrans (Amson et al. 2017a). But mean diaphysis is higher than that of most extant values of other trabecular parameters of interest are xenarthrans, falling in the upper distribution of also presented in Table 3. For the humeral head, armadillos and extant sloths (Fig. 3C). One should using a ROI representing 72% of the maximum note, however, that this parameter does not yield any volume (see Material and Methods section), clear distinction among lifestyles. The humeral CSS armadillos are distinguished from other extant at mid-diaphysis of Scelidotherium, on the other xenarthrans by their high values (i.e. more anisotropic hand, falls among anteaters (Fig. 3D). architecture). Both Hapalops and Lestodon (n=1 in There is a clear tendency for the radial diaphysis each case) fall in the upper distribution (i.e., more GC to be higher in armadillos, intermediate in anisotropic) of extant sloths and anteaters (Fig. 4A). anteaters, and lower in extant sloths. Hapalops (n=1; The same pattern is found for the full ROI in Lestodon

Table 3. Mean values of trabecular parameters of interest for each lifestyle category and extinct taxon.

-3 -1 DA (NU) Conn.D (nb.mm ) Tb.Th (mm) Tb.Sp (mm) BS/TV (mm ) BV/TV (NU) Humeral head 100% Armadillos 0.60 12.35 0.25 0.47 3.38 0.41 Anteaters 0.40 11.59 0.26 0.41 3.43 0.45 Extant sloths 0.43 9.36 0.31 0.49 3.14 0.44 Lestodon 0.50 0.58 0.80 0.81 1.23 0.58 Humeral head 72% Armadillos 0.62 12.31 0.24 0.46 3.57 0.41 Anteaters 0.40 11.38 0.26 0.42 3.56 0.45 Extant sloths 0.44 9.03 0.32 0.50 3.16 0.45 Hapalops 0.52 3.22 0.39 0.56 3.39 0.50 Radial trochlea 100% Armadillos 0.79 16.05 0.34 0.37 3.87 0.49 Anteaters 0.63 11.19 0.30 0.44 3.20 0.44 Extant sloths 0.56 8.75 0.30 0.52 2.39 0.40 Hapalops 0.60 3.03 0.26 0.85 1.55 0.24 Glossotherium 0.43 1.02 0.43 1.25 0.59 0.23

Footnotes. Percentage indicates the cropping coefficient that was used (100% denoting the lack thereof; see Material and Methods section). Abbreviation: NU, no units.

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 10 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

(no data for Hapalops, see Material ad Methods provide the canonical coefficients (weights) of each section). In the distal radius (trochlea), the trabecular explanatory variable for each analysis in SOM 5. For architecture of armadillos is again found as more Hapalops, 18 parameters could be initially included in anisotropic than in the other extant categories. the analysis (diaphyseal and trabecular parameters, Moreover, the main distribution of extant sloths is from both the humerus and radius). Due to high found as clustering at the level of the lower values of correlation among some variables (Conn.D between anteaters. The DA value of Hapalops falls above the two ROIs; between Tb.Th and Tb.Sp of both ROIs; main distribution of extant sloths, within that of between BS and BV of the radial trochlea ROI), four anteaters (Fig. 4B). Glossotherium is the sampled variables were excluded (see list of included variables taxon with the single lowest DA value (most isotropic in SOM 5. The recovered optimal Lambda is 0 (no structure). One should note, however, that DA was significant correlation of the trait values with significantly correlated to body size (see Discussion). phylogeny) and the discrimination is optimal (training The main direction of the trabeculae (MDT) in the misclassification error of 0%). Hapalops is classified radial trochlea (humeral head did not yield lifestyle in the category of extant sloths’ lifestyle with a high discrimination; Amson et al. 2017b) of both Hapalops posterior probability (>99%). Indeed, it falls close to and Glossotherium falls outside the distribution of extant sloths’ distribution along the Discriminant Axis extant xenarthrans (Fig. 4C). In both cases, the MDT (pDA) 1 (Fig. 5A). However, Hapalops clearly falls falls closer to the distribution of extant sloths. beyond the distribution of extant xenarthrans along pDA2. The parameter contributing the most to the

discrimination is the DA (that of the radial trochlea for pDA1 and that of the humeral head for pDA2; see Phylogenetically flexible discriminant SOM 5). analyses For Lestodon, eight parameters could be included Each studied “ground sloth” was subject to an (from the radial diaphysis and humeral head independent analysis (see Material and Methods), to trabeculae), of which one was excluded because of predict the most probable lifestyle among the three collinearity (present between Tb.Th and Tb.Sp). The broad lifestyle categories represented by armadillos, recovered optimal Lambda is 0.84, and training anteaters, and extant sloths, respectively. The results misclassification error is 50%. It is classified in the regarding classification of each “ground sloth” are armadillos’ lifestyle category with a rather low given in Table 4, and the corresponding outcomes of posterior probability (64%), the second most probable the training data (posterior probability of the classification being to anteaters (35%). According to classification of the extant species according to each this analysis, a classification in extant sloth’s category discriminant analysis) are given in SOM 4. We also is very improbable (0.006%). Lestodon falls beyond

Figure 4. Univariate comparisons of trabecular anisotropy parameters. A, degree of anisotropy (DA) in the humeral head ROI, reduced at 72% of its maximum size (see Material and Methods section); B, DA in the radial trochlea; C, main direction of the trabeculae (MDT) in the radial trochlea. Abbreviations: ant, anteaters; arma, armadillos; Glos, Glossotherium; Hapa, Hapalops; sloth, extant sloths.

n=22 n=11 n=6 n=1 n=1 Anterior A C (0;0) arma ant sloth Hapa Glos Hp72 DA (raw) Hp72DA 0.3 0.6

arma ant sloth Hapa Lest Medial (270;0) B n=22 n=9 n=5 n=1 n=1

Rd DA (raw) RdDA Distal: 0.5 0.7 0.9

arma ant sloth Hapa Glos

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 11 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

the distribution of extant xenarthrans (Fig. 5B). The The four discriminant analyses we performed vary parameter contributing the most to the discrimination greatly in the number of included parameters. As is the ‘size-normalized’ Tb.Th (for both pDA1 and expected, analyses including more parameters pDA2). yielded a better discrimination, i.e., a lower misclassification error. The lowest misclassification For Glossotherium, eight parameters could be error (0%) was obtained for the analysis of Hapalops, included (from the radial diaphysis and trabeculae of for which it was possible to include 14 parameters (18 the radial trochlea). The recovered optimal Lambda is before exclusion of collinear parameters) from both 0.88, and training misclassification error is 35%. The the diaphysis and epiphyseal trabeculae. The worst most probable classification is to anteaters (50%), discrimination (69% of misclassification error) was followed by the equally probable classifications to found for the analysis of Scelidotherium, for which armadillos or extant sloths (each 25%). only two parameters, from the humeral diaphysis, Glossotherium falls within the distribution of extant could have been included. This lends support to the xenarthrans, but outside the distribution of each approach of combining parameters from several bone lifestyle class, just outside that of anteaters (Fig. 5C). compartments, if one endeavours to discriminate The parameters contributing the most to the lifestyles based on these parameters. discrimination are the DA (pDA1) and ‘size- normalized’ BS (pDA2). Several of the investigated parameters were significantly correlated with body size. To attempt to For Scelidotherium, only two parameters could be prevent the size of the studied taxa from influencing included (from the humeral diaphysis). An optimal the analysis, a common approach is to size-correct Lambda of 0.96 and a high training misclassification the raw data using the residuals of a regression of the error of 69% were recovered. The three possible trait against a body size proxy (Mccoy et al. 2006). classifications are roughly equally probable (anteater: This proved to be challenging for extinct sloths, 37%; extant sloth: 36%; armadillo: 27%). because, for most of them, body size largely exceeds Scelidotherium basically falls in the middle of the that of extant xenarthrans (Vizcaíno et al. 2017). This distribution of extant xenarthrans (Fig. 5D). The potentially makes the size regressions spurious, as parameter contributing the most to the discrimination the extreme values over-influence the regression is CSS (for both pDA1 and pDA2). coefficients. This is not a trivial consideration for our dataset. For instance, if one would size-correct the DA in the radial trochlea using the residuals of the corresponding size regression, the medium-sized DISCUSSION extinct sloth Glossotherium, of which the raw DA value was found as the lowest of the dataset, would On the whole, the classification of extinct sloths to fall in the middle of the overall distribution. For those one of the extant xenarthran lifestyles (that of parameters that are dimensionless, we hence armadillos, anteaters, or extant sloths) based on decided to use the untransformed data. But this is forelimb bone structure proved to be challenging. likely to be biased as well, due the potential presence This appears to be due to at least three obvious of allometry. For instance, the scaling exponent of the causes: (1) the imperfect lifestyle discrimination degree of anisotropy (DA) across primates in the based on diaphyseal and trabecular parameters, (2) humeral and femoral head was found by Ryan & the difficulties raised by the size correction (for some Shaw (2013) to be significantly negative (but close to parameters), and (3) the fact that the values of extinct 0, which would have denoted isometry). We also taxa are outliers with respect to the distribution of found a negative scaling exponent for one of the extant xenarthrans (for some parameters). investigated ROI, the radial trochlea. It would be suboptimal to exclude this parameter, especially because it was found as the best functionally

Predicted class P(ant) P(arma) P(sloth) Table 4. Lifestyle classification of the extinct taxa as predicted by Hapalops sloth 0.00 0.00 1.00 phylogenetically flexible Lestodon armatus arma 0.35 0.64 0.01 discriminant analyses (because of the difference in the included Glossotherium robustum ant 0.50 0.25 0.25 predictive variables for each 0.37 0.27 0.36 taxon, a dedicated discriminant Scelidotherium leptocephalum ant analysis was performed for each of them). Footnotes. Abbreviations: P(“class”), the posterior probability for the extinct taxon to be classified as “class”; ant, anteater; arma, armadillo; sloth, extant sloth.

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 12 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

A B Figure 5. Phylogenetically ant flexible linear discriminant arma analyses using humeral and sloth unknown radial bone structure parameters. One analysis per extinct taxon (referred as of − 1 0 1 2

“unknown” class) was − 2 performed, because of the pDA 2 (40.7%) pDA − 3 difference in the parameters 2 (13.09%) pDA that could be included (see − 10 0 10 20 30 Material and Methods section −30 −20 −10 0 10 20 −3 −2 −1 0 1 2 3 and Table 1). A, Hapalops; B, Lestodon; C, Glossotherium; D, pDA 1 (59.3%) pDA 1 (86.91%) Scelidotherium. Abbreviations: ant, anteaters; arma, C D armadillos; sloth, extant sloths. Next to each discriminant axis is given between brackets the corresponding percentage of explained between-group variance. The size of extinct − 1.0 0.0 sloths’ representations gives a − 1.0 0.0 rough indication of their body

sizes. 2 (10.22%) pDA 2 (13.16%) pDA − 2.0 − 2.0

−2 −1 0 1 −2 −1 0 1 pDA 1 (89.78%) pDA 1 (86.84%) discriminating parameter in extant xenarthrans diaphysis of this taxon seem to be notably affected by (Amson et al. 2017a). It was also singled out as bone mass increase. Finding a compact humerus is reflecting joint loading in primates better than other particularly surprising, as the stylopod can be parameters (Tsegai et al. 2018), and, more generally, expected to be less compact than the zeugopod in DA was found as functionally informative in several terrestrial mammals (Amson & Kolb 2016). In the analyses about that clade (e.g. Ryan & Ketcham case of Lestodon, it was not obvious from univariate 2002; Griffin et al. 2010; Barak et al. 2013; Su et al. comparisons that its bone structure was outlying, but 2013; Georgiou et al. 2018; Ryan et al. 2018; Tsegai both the latter and Hapalops fell outside the range of et al. 2018). A tendency for a more anisotropic extant xenarthrans in the respective discriminant structure in the femoral head of arboreal squirrels was analyses. One may hence conclude that, based on also demonstrated (Mielke et al. 2018b). A way to their bone structure, the humerus and radius of both improve accuracy of the size-correction using Hapalops and Lestodon were likely involved in a residuals of a regression against a body size proxy loading regime different from those associated with would be, in our case, to include to the sampling the lifestyles of extant xenarthrans. For Hapalops, xenarthrans that have a body size between that of one can however note that the phylogenetically extant species and that of the giant “ground sloths”, informed discriminant analysis strongly supports a i.e., with a mass roughly between 50 kg and 300 kg. classification within extant sloths’ category, which Unfortunately, the number of known xenarthrans of might indicate that some aspects of their mechanical this size range is very limited. environment were similar. The main direction of the trabecular (MDT) also agrees with the fact that the It was already obvious from univariate bone structure of extant sloths is different from that of comparisons that the bone structure in Hapalops, the Hapalops, but that the former represent the most small-sized extinct sloth, departed from the condition similar of the three extant lifestyles discriminated here observed in extant xenarthrans. Indeed, the overall (Fig. 4C). Based on bone gross morphology, great compactness of its humeral diaphysis does not Hapalops was previously reconstructed as partly or seem to be matched by any other sampled primarily arboreal (Matthew 1912; White 1997). Both xenarthran (but see aquatic specialization of bone structure and gross morphology therefore seem Thalassocnus; Amson et al. 2014). This does not to point in the same direction for the reconstruction of seem to be a systemic bone mass increase (Amson Hapalops’ lifestyle. The large-sized Lestodon, on the et al. 2018), because neither the trabecular other hand, is not classified with strong support to one parameters nor the compactness of the radial

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 13 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

of the extant groups. The least probable classification homeostasis and/or metabolism (Eleazer & is to extant sloths’ lifestyle (0.03% of posterior Jankauskas 2016; and references therein). While probability), which might suggest that the bone more experimental data is required to discuss it structure of Lestodon resembles more that of beyond speculation, it was reported that extant sloths anteaters and armadillos. Naturally, suspensory (at least the two-toed sloth Choloepus) are prone to posture has never been purported for this elephant- soft tissue mineralization likely due to mineral sized sloth. Lestodon was interpreted as traviportal imbalance (Han & Garner 2016). One can therefore (slow-moving with both quadrupedal and bipedal speculate that the observed great quantity of stances) by Toledo (1996), and the forelimb gross diaphyseal trabeculae might be a storage mechanism morphology was found to be consistent with fossorial for mineral in excess. activity (but probably not to procure food (Coombs The extremely low metabolism of extant sloths 1983); see Bargo et al. (2000) for a more tempered was suggested by Montañez-Rivera et al. (2018) as interpretation). Including other fossorial and non- a potential explaining factor for their low cortical fossorial taxa in the sampling of the bone structure compactness (CC). Indeed, they found that extant analysis will be necessary to suggest a more precise sloths depart in that regard from other extant assertion regarding the digging habits of this taxon xenarthrans as well as from two extinct sloths (the (but its large size might be problematic, see above). small-sized Hapalops and Parocnus). No quantitative The two other extinct sloths subject to a discriminant assessment of CC was performed here. But we can analysis, Glossotherium and Scelidotherium, differ report that, at mid-diaphysis, the CC of the sampled from the former two in falling within the distribution of extinct sloths was generally observed as low (when extant xenarthrans. However, in neither case is the an observation was possible), similar to armadillos classification clear, and it seems that acquiring and anteaters. Nevertheless, two specimens showed additional bone structure parameters will be a rather porous cortex, Hapalops (humerus; necessary to draw reliable conclusions. MNHN.F.SCZ162) and Glossotherium (radius; The Mc III did not yield clear discrimination among MNHN.F.PAM756), though not as porous as that of the extant lifestyles and was hence not included in the most extant sloths. A dedicated analysis of extinct discriminant analyses. But one can note that an sloths’ CC is required to investigate this trait and interesting pattern was observed in the cross- possibly use it to inform metabolic rate reconstruction sectional shape (CSS) of extinct sloths at mid- in extinct sloths. diaphysis. Indeed, high values, denoting elliptic sections, are found in Valgipes and Glossotherium. Comparison of long bone’s cross-sections among Such a bone structure is expected to be suited to specimens should be performed at the same location, resist bending along its major axis (Ruff & Hayes usually defined as a percentage of the bone’s length 1983). This is consistent with previous lifestyle (e.g., Ruff & Hayes 1983). Here, mid-diaphysis (i.e., reconstruction of Glossotherium, which is argued to 50% of bone length) was selected for complete have had fossorial habits (Coombs 1983; Bargo et al. bones, and, for fragmentary specimens (some 2000) supposedly entailing a well-marked main fossils), it is the preserved level closest to mid- direction of bending. Furthermore, it might suggest diaphysis that was used (the other specimens were that Valgipes had similar habits, which, to our resampled accordingly). Because of the xenarthran knowledge, was never purported. bones’ morphology, most examined cross-sections were located at the level of a prominent bony process. A medulla filled with spongy bone was observed One could therefore consider selecting cross- in large-sized mammals, and argued to be a potential sections avoiding those processes to test their adaptation to graviportality (Houssaye et al. 2015). It influence on bone structural parameters. Acquiring does not seem to be possible to easily draw such a cross-sectional properties along the whole diaphysis conclusion for xenarthrans: whatever their lifestyle, and assessing the proximodistal evolution of xenarthrans with a mass of roughly 5 kg (e.g., biomechanical properties can also be considered for Tamandua) and over tend to fill their medullary cavity complete bones (Houssaye & Botton-Divet 2018). with spongy bone. This is true for the forelimb, as described here (and as also reported by Houssaye et al. (2015) for the humerus), but likely also for the hind In addition to lifestyle, one can expect that the limb: a ‘naturally sectioned’ tibia of the small-sized factors affecting bone structure are the individual’s Nothrotherium (less than ca. 100 kg; Amson et al. age, health status, and possibly other features 2016) reveals that the medullary cavity is entirely varying intraspecifically (such as sex differences; filled with dense spongy bone (Fig. 2D). In the case Eckstein et al. 2007). Details regarding these of xenarthrans, the great quantity of diaphyseal potential factors are mostly unknown for fossils (and often for recent specimens as well). To control for trabeculae might be related to another aspect affecting bone structure, such as mineral these factors as much as feasible, the sampled specimens were chosen to be devoid of apparent

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 14 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

bone diseases and skeletally mature (even though Cartelle (Museu de Ciencias Naturais da Pontifícia several presented a remnant of epiphyseal line, see Universidade Católica de Minas Gerais, Belo above). It is our assumption that variations in bone Horizonte, Brazil), Anneke H. van Heteren structure that relate to a different lifestyle can be (Zoologische Staatssammlung München), Frieder expected to be of greater magnitude than Mayer and Christiane Funk (Museum für Naturkunde intraspecific variations. But this chiefly remains to be Berlin), Thomas Kaiser and Nelson Ribeiro demonstrated. Mascarenhas (Universität Hamburg), Irina Ruf and Katrin Krohmann (Senckenberg Forschungsinstitut und Naturmuseum), Stefan Merker (Staatlichen Museums für Naturkunde Stuttgart), Eva Bärmann (Zoologische Forschungsmuseum Alexander CONCLUSION Koenig). Anneke H. van Heteren (Zoologische Staatssammlung München), Patrick Arnold Bone structure of the diaphysis and epiphyses of (Friedrich-Schiller-Universität Jena), and Aurore the third metacarpal, humerus, and radius was here Canoville (NC State University) are acknowledged for investigated in several species of extinct sloths, helping with acquisition of the extant species scans. comparing it to that of extant xenarthrans. Related We thank Patricia Wills, Marta Bellato, and Maïté parameters were successfully acquired and included Adam (AST-RX platform, MNHN) for acquiring the in phylogenetically flexible discriminant analyses. The extinct species scans. We acknowledge Luis D. latter constitute, to our knowledge, the first analyses Verde Arregoitia for his help with the function pFDA. that conjointly include both diaphyseal and trabeculae Andrew Pitsillides (acting as a reviewer), an parameters to discriminate lifestyles. However, no anonymous reviewer, Alexandra Houssaye (acting as extinct sloths are here confidently ascribed to one of a PCI Paleo Recommender) and an additional PCI the lifestyles exhibited by extant xenarthrans. This Paleo Recommender are thanked for the might be due to several factors, and we identified as improvement they brought to the manuscript. challenges for the present analysis the lack of discrimination power of some parameters, the difficulties raised by size-correlated parameters, and the fact that some parameters fall outside the range described by extant taxa. The humeral and radial ADDITIONAL INFORMATION structure of the small-sized Hapalops, from the Miocene of Argentina, was nevertheless found as Funding more reminiscent of that of extant sloths, which agrees with the conclusions drawn based on gross This research received support from the morphology. The humeral and radial structure of the SYNTHESYS Project http://www.SYNTHESYS.info/ large-sized Lestodon, from the Pleistocene of which is financed by European Community Research Argentina, clearly departs from that of extant sloths, Infrastructure Action under the FP7 Integrating and is more similar to that of anteaters and Activities Programme. EA was funded by the armadillos. The singular bone structure of Alexander von Humboldt Foundation. JAN and EA xenarthrans, including a medullary cavity filled with were funded by the German Research Council (DFG spongy bone in most taxa, and a low cortical EXC 1027 and DFG AM 517/1-1, respectively). compactness in extant sloths, deserves further investigation. Because Xenarthra is argued to be one Competing interests of the four early diverging clades of placental The authors declare they have no personal or mammals (Delsuc & Douzery 2008; Asher et al. 2009; financial conflict of interest relating to the content of Gaudin & Croft 2015), such investigations are not this preprint. EA is a Recommender for PCI Paleo only important for the understanding of the evolutionary history of the clade, but potentially for that of Mammalia as well. Author contributions Conceptualization and methodology, EA, JAN; Formal analysis, EA; Investigation, EA; Writing – Original draft, EA; Writing – Review & editing, EA, AKNOWLEDGEMENTS JAN. All authors gave final approval for publication. Data availability We warmly thank the following curators and collection managers: Guillaume Billet (Muséum All the raw scans of fossil specimens sampled for national d’Histoire naturelle, Paris; MNHN), Cástor the present analysis will be available from the MNHN

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 15 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

collection database pending an embargo. The extant Amson, E., J. D. Carrillo, and C. Jaramillo. 2016. species specimens sampled come from various Neogene sloth assemblages (Mammalia, collections. The corresponding raw scans are Pilosa) of the Cocinetas Basin (La Guajira, available upon reasonable request to the authors. Colombia): implications for the Great American Biotic Interchange. Palaeontology 59. Supplementary information Amson, E., and C. Kolb. 2016. Scaling effect on the SOM 1. Raw data. Excel document, of which each mid-diaphysis properties of long bones—the worksheet corresponds to a sampled region. case of the Cervidae (deer). The Science of SOM 2. ImageJ macro to crop isometrically a stack Nature 103:58. in 3D. Amson, E., C. de Muizon, and T. J. Gaudin. 2017b. A SOM 3. ImageJ macro to acquire mid-diaphyseal reappraisal of the phylogeny of the Megatheria parameters. (Mammalia: Tardigrada), with an emphasis on SOM 4. Lifestyle classification of the extant taxa, the the relationships of the Thalassocninae, the training data of the phylogenetically flexible marine sloths. Zoological Journal of the Linnean discriminant analyses (one analysis was performed Society 179:217–236. per extinct taxon, each on a different worksheet). Amson, E., C. de Muizon, M. Laurin, C. Argot, and V. SOM 5. Canonical coefficients for each de Buffrénil. 2014. Gradual adaptation of bone phylogenetically flexible discriminant analysis (one structure to aquatic lifestyle in extinct sloths analysis was performed per extinct taxon, each on a from Peru. Proceedings of the Royal Society B: different worksheet). Biological Sciences 281:20140192.

Amson, E., and J. A. Nyakatura. 2017. The postcranial musculoskeletal system of Xenarthrans: Insights from over two centuries of REFERENCES research and future directions. Journal of Mammalian Evolution, doi: 10.1007/s10914- Abba, A. M., and M. Superina. 2016. Dasypus 017-9408-7. hybridus (Cingulata: Dasypodidae). Mammalian Species 48:10–20. Asher, R. J., N. Bennett, and T. Lehmann. 2009. The new framework for understanding placental Amson, E., C. Argot, H. G. McDonald, and C. de mammal evolution. BioEssays 31:853–864. Muizon. 2015a. Osteology and functional morphology of the hind limb of the marine sloth Barak, M. M., D. E. Lieberman, and J.-J. Hublin. Thalassocnus (Mammalia, Tardigrada). Journal 2011. A Wolff in sheep’s clothing: Trabecular of Mammalian Evolution 22:355–419. bone adaptation in response to changes in joint loading orientation. Bone 49:1141–1151. Amson, E., C. Argot, H. G. McDonald, and C. de Muizon. 2015b. Osteology and functional Barak, M. M., D. E. Lieberman, D. Raichlen, H. morphology of the axial postcranium of the Pontzer, A. G. Warrener, and J. J. Hublin. 2013. marine sloth Thalassocnus (Mammalia, Trabecular evidence for a human-like gait in Tardigrada) with paleobiological implications. Australopithecus africanus. PLoS ONE 8. Journal of Mammalian Evolution 22:473–518. Bargo, M. S., N. Toledo, and S. F. Vizcaíno. 2012. Amson, E., P. Arnold, A. H. van Heteren, A. Canoville, Paleobiology of the Santacrucian sloths and and J. A. Nyakatura. 2017a. Trabecular anteaters (Xenarthra, Pilosa). Pp. 216–242 in S. architecture in the forelimb epiphyses of extant F. Vizcaíano, R. F. Kay, and M. S. Bargo, eds. xenarthrans (Mammalia). Frontiers in Zoology Early Miocene Paleobiology in Patagonia: High- 14:52. Latitude Paleocommunities of the Santa Cruz Formation. Cambridge University Press, Amson, E., G. Billet, and C. de Muizon. 2018. Cambridge. Evolutionary adaptation to aquatic lifestyle in extinct sloths can lead to systemic alteration of Bargo, M. S., and S. F. Vizcaíno. 2008. Paleobiology bone structure. Proceedings of the Royal of Pleistocene ground sloths (Xenarthra, Society B 285:20180270. Tardigrada): biomechanics, morphogeometry and ecomorphology applied to the masticatory

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 16 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

apparatus. Ameghiniana 45:175–196. Gainesville.

Bargo, M. S., S. F. Vizcaíno, F. M. Archuby, and R. Delsuc, F., M. Kuch, G. C. Gibb, J. Hughes, P. Szpak, E. Blanco. 2000. Limb bone proportions, J. Southon, J. Enk, A. T. Duggan, and H. N. strength and digging in some Lujanian (late Poinar. 2018. Resolving the phylogenetic Pleistocene-early Holocene) mylodontid ground position of Darwin’s extinct ground sloth sloths (Mammalia, Xenarthra). Journal of (Mylodon darwinii) using mitogenomic and Vertebrate Paleontology 20:601–610. nuclear exon data. Proceedings of the Royal Society B 285:20180214. Biewener, A. A., N. L. Fazzalari, D. D. Konieczynski, and R. V. Baudinette. 1996. Adaptive changes DeSilva, J. M., and M. J. Devlin. 2012. A comparative in trabecular architecture in relation to functional study of the trabecular bony architecture of the strain patterns and disuse. Bone 19:1–8. talus in humans, non-human primates, and Australopithecus. Journal of Human Evolution Carlini, A. A., R. Pascual, M. A. Reguero, G. J. 63:536–551. Scillato-Yané, E. P. Tonni, and S. F. Vizcaíno. 1990. The first Paleogene land placental Doube, M., M. M. Kłosowski, I. Arganda-Carreras, F. mammal from Antarctica: its paleoclimatic and P. Cordelières, R. P. Dougherty, J. S. Jackson, paleobrogeographical bearings. P. 325 in B. B. Schmid, J. R. Hutchinson, and S. J. Cox and J. Reveal, eds. Abstracts IV Shefelbine. 2010. BoneJ: Free and extensible International Congress of Systematic and bone image analysis in ImageJ. Bone 47:1076– Evolutionary Biology. University of Maryland, 1079. Baltimore, MD. Eckstein, F., M. Matsuura, V. Kuhn, M. Priemel, R. Cartelle, C., G. De Iuliis, and R. L. Ferreira. 2009. Müller, T. M. Link, and E. M. Lochmüller. 2007. Systematic revision of tropical Brazilian Sex differences of human trabecular bone scelidotheriine sloths (Xenarthra, microstructure in aging are site-dependent. Mylodontoidea). Journal of Vertebrate Journal of Bone and Mineral Research 22:817– Paleontology 29:555–566. 824.

Carter, D. R., and G. S. Beaupré. 2001. Skeletal Eleazer, C. D., and R. Jankauskas. 2016. Mechanical Function and form. Cambridge University Press, and metabolic interactions in cortical bone Cambridge. development. American Journal of Physical Anthropology 160:317–333. Coombs, M. C. 1983. Large mammalian clawed herbivores: a comparative study. Transactions Fariña, R. A., and R. E. Blanco. 1996. Megatherium, of the American Philosophical Society, New the stabber. Proceedings of the Royal Society B series 73:1–96. 263:1725–1729.

Coutier, F., L. Hautier, R. Cornette, E. Amson, and G. Fariña, R. A., S. F. Vizcaíno, and M. S. Bargo. 1998. Billet. 2017. Orientation of the lateral Body mass estimations in Lujanian (late semicircular canal in Xenarthra and its links with Pleistocene-early Holocene of South America) head posture and phylogeny. Journal of mammal megafauna. Mastozoología Morphology 278. Neotropical 5:87–108.

Delsuc, F., F. M. Catzeflis, M. J. Stanhope, and E. J. Flynn, J. J., and C. C. Swisher III. 1995. Cenozoic Douzery. 2001. The evolution of armadillos, South American land mammal ages: correlation anteaters and sloths depicted by nuclear and to global geochronologies. Geochronology Time mitochondrial phylogenies: implications for the Scales and Global Stratigraphic Correlation, status of the enigmatic fossil Eurotamandua. SEPM Special Publication 54:317–333. Proceedings of the Royal Society B 268:1605– 1615. Gaudin, T. J. 2004. Phylogenetic relationships among sloths (Mammalia, Xenarthra, Tardigrada): the Delsuc, F., and E. J. P. Douzery. 2008. Recent craniodental evidence. Zoological Journal of the advances and future prospects in xenarthran Linnean Society 140:255–305. molecular phylogenetics. Pp. 11–23 in S. F. Vizcaino and W. J. Loughry, eds. The Biology of Gaudin, T. J., and D. A. Croft. 2015. Paleogene the Xenarthra. University Press of Florida, Xenarthra and the evolution of South American

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 17 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

mammals. Journal of Mammalogy 96:622–634. the flexor sheath ridge. Journal of Human Evolution 67:60–75. Georgiou, L., T. L. Kivell, D. H. Pahr, and M. M. Skinner. 2018. Trabecular bone patterning in Kay, R. F., B. J. Macfadden, R. H. Madden, H. the hominoid distal femur. PeerJ 6:e5156. Sandeman, and F. Anaya. 1998. Revised age of the salla beds, Bolivia, and its bearing on the Germain, D., and M. Laurin. 2005. Microanatomy of age of the deseadan South American land the radius and lifestyle in amniotes (Vertebrata, mammal “age.” Journal of Vertebrate Tetrapoda). Zoologica Scripta 34:335–350. Paleontology 18:189–199.

Gibb, G. C., F. L. Condamine, M. Kuch, J. Enk, N. Kivell, T. L. 2016. A review of trabecular bone Moraes-Barros, M. Superina, H. N. Poinar, and functional adaptation: what have we learned F. Delsuc. 2016. Shotgun mitogenomics from trabecular analyses in extant hominoids provides a reference phylogenetic framework and what can we apply to fossils? Journal of and timescale for living xenarthrans. Molecular Anatomy 228:569–594. Biology and Evolution 33:621–642. Kivell, T. L., M. M. Skinner, R. Lazenby, and J.-J. Griffin, N. L., K. D’Août, T. M. Ryan, B. G. Richmond, Hublin. 2011. Methodological considerations for R. A. Ketcham, and A. Postnov. 2010. analyzing trabecular architecture: an example Comparative forefoot trabecular bone from the primate hand. Journal of anatomy architecture in extant hominids. Journal of 218:209–25. human evolution 59:202–13. Kolb, C., T. M. Scheyer, K. Veitschegger, A. M. Gross, T., T. L. Kivell, M. M. Skinner, N. H. Nguyen, Forasiepi, E. Amson, A. A. E. Van der Geer, L. and D. H. Pahr. 2014. A CT-image-based W. Van den Hoek Ostende, S. Hayashi, and M. framework for the holistic analysis of cortical R. Sánchez-Villagra. 2015. Mammalian bone and trabecular bone morphology. palaeohistology: A survey and new data with Palaeontologia Electronica 17:1–13. emphasis on island forms. PeerJ 2015.

Hammer, A. 2010. The structure of the femoral neck: Lazenby, R. A., D. M. L. Cooper, S. Angus, and B. A physical dissection with emphasis on the Hallgrímsson. 2008. Articular constraint, internal trabecular system. Annals of Anatomy handedness, and directional asymmetry in the 192:168–177. human second metacarpal. Journal of Human Evolution 54:875–885. Han, S., and M. M. Garner. 2016. Soft tissue mineralization in captive 2-toed sloths. Matthew, W. D. 1912. The ancestry of edentates: as Veterinary Pathology 53:659–665. illustrated by the skeleton of Hapalops, a tertiary ancestor of the ground sloths. American Hayssen, V. 2010. Bradypus variegatus (Pilosa: Museum Journal 12:300–303. Bradypodidae). Mammalian Species 42:19–32. Mccoy, M. W., B. M. Bolker, C. W. Osenberg, B. G. Houssaye, A., and L. Botton-Divet. 2018. From land Miner, and J. R. Vonesh. 2006. Size correction: to water: evolutionary changes in long bone Comparing morphological traits among microanatomy of otters (Mammalia: populations and environments. Oecologia Mustelidae). Biological Journal of the Linnean 148:547–554. Society, doi: 10.1093/biolinnean/bly118. Meier, P. S., C. Bickelmann, T. M. Scheyer, D. Houssaye, A., K. Waskow, S. Hayashi, A. H. Lee, and Koyabu, and M. R. Sánchez-Villagra. 2013. J. R. Hutchinson. 2015. Biomechanical Evolution of bone compactness in extant and evolution of solid bones in large animals: a extinct moles (Talpidae): exploring humeral microanatomical investigation. Biological microstructure in small fossorial mammals. Journal of the Linnean Society 117:350–371. BMC evolutionary biology 13:55.

Huynh Nguyen, N., D. H. Pahr, T. Gross, M. M. Mielke, F., E. Amson, and J. A. Nyakatura. 2018a. Skinner, and T. L. Kivell. 2014. Micro-finite Morpho-functional analysis using procrustes element (µFE) modeling of the siamang superimposition by static reference. (Symphalangus syndactylus) third proximal Evolutionary Biology, doi: 10.1007/s11692-018- phalanx: The functional role of curvature and 9456-9.

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 18 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

Mielke, M., J. Wölfer, A. H. van Heteren, E. Amson, evolution of sloths. Journal of Mammalian and J. A. Nyakatura. 2018b. Trabecular Evolution 19:159–169. architecture in the sciuromorph femoral head: allometry and functional adaptation. Zoological Ruff, C. B., and W. C. Hayes. 1983. Cross-sectional Letters 4:10. geometry of Pecos Pueblo femora and tibiae - a biomechanical investigation: I. Method and Montañez-Rivera, I., J. A. Nyakatura, and E. Amson. general patterns of variation. American Journal 2018. Bone cortical compactness in “tree sloths” of Physical Anthropology 60:359–81. reflects convergent evolution. Journal of Anatomy, doi: 10.1111/joa.12873. Ruff, C. B., B. Holt, and E. Trinkaus. 2006. Who’s afraid of the big bad Wolff?: “Wolff’s law” and Motani, R., and L. Schmitz. 2011. Phylogenetic bone functional adaptation. American Journal of versus functional signals in the evolution of Physical Anthropology 129:484–498. form–function relationships in terrestrial vision. Evolution 65:2245–2257. Ryan, T. M., K. J. Carlson, A. D. Gordon, N. Jablonski, C. N. Shaw, and J. T. Stock. 2018. Nyakatura, J. A. 2012. The convergent evolution of Human-like hip joint loading in Australopithecus suspensory posture and locomotion in tree africanus and Paranthropus robustus. Journal sloths. Journal of Mammalian Evolution 19:225– of Human Evolution 121:12–24. 234. Ryan, T. M., and R. A. Ketcham. 2002. The three- Owen, R. 1839. Part I. Fossil Mammalia. Pp. 1–111 dimensional structure of trabecular bone in the in C. Darwin, ed. The Zoology of the Voyage of femoral head of strepsirrhine primates. Journal HMS Beagle. Smith Elder and Co, London. of Human Evolution 43:1–26.

Owen, R. 1861. Memoir on the Megatherium, or giant Ryan, T. M., and C. N. Shaw. 2013. Trabecular bone ground-sloth of America (Megatherium microstructure scales allometrically in the americanum, Cuvier). Williams and Norgate, primate humerus and femur. Proceedings of the London. Royal Society B 280:20130172.

Pagel, M. 1999. Inferring the historical patterns of Schindelin, J., I. Arganda-Carreras, E. Frise, V. biological evolution. Nature 401:877–884. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J.-Y. Tinevez, Patiño, S. J., and R. A. Fariña. 2017. Ungual D. J. White, V. Hartenstein, K. Eliceiri, P. phalanges analysis in Pleistocene ground sloths Tomancak, and A. Cardona. 2012. Fiji: an open- (Xenarthra, Folivora). Historical Biology source platform for biological-image analysis. 29:1065–1075. Nature Methods 9:676–682.

Perkins, M. E., J. G. Fleagle, M. T. Heizler, B. Nash, Schindelin, J., C. T. Rueden, M. C. Hiner, and K. W. T. M. Bown, A. A. Tauber, and M. T. Dozo. Eliceiri. 2015. The ImageJ ecosystem: An open 2012. Tephrochronology of the Miocene Santa platform for biomedical image analysis. Cruz and Pinturas Formations, Argentina. Pp. Molecular Reproduction and Development 23–40 in S. F. Vizcaíno, R. F. Kay, and M. S. 82:518–529. Bargo, eds. Early Miocene Paleobiology in Patagonia: high-latitude paleocommunities of Schneider, C. A., W. S. Rasband, and K. W. Eliceiri. the Santa Cruz Formation. Cambridge 2012. NIH Image to ImageJ: 25 years of image University Press, Cambridge. analysis. Nature Methods 9:671–675.

Pontzer, H., D. E. Lieberman, E. Momin, M. J. Devlin, Shaw, C. N., and T. M. Ryan. 2012. Does skeletal J. D. Polk, B. Hallgrímsson, and D. M. L. anatomy reflect adaptation to locomotor Cooper. 2006. Trabecular bone in the bird knee patterns? Cortical and trabecular architecture in responds with high sensitivity to changes in load human and nonhuman anthropoids. American orientation. The Journal of Experimental Biology journal of physical anthropology 147:187–200. 209:57–65. Skinner, M. M., N. B. Stephens, Z. J. Tsegai, A. C. Pujos, F., T. J. Gaudin, G. De Iuliis, and C. Cartelle. Foote, N. H. Nguyen, T. Gross, D. H. Pahr, J. 2012. Recent advances on variability, morpho- Hublin, and T. L. Kivell. 2015. Human-like hand functional adaptations, dental terminology, and use in Australopithecus africanus. Science

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 19 bioRxiv preprint doi: https://doi.org/10.1101/318121; this version posted September 21, 2018. The copyright holder for this preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under aCC-BY-NC-ND 4.0 International license.

PREPRINT

347:395–399. 41:D1027–D1033.

Slater, G. J., P. Cui, A. M. Forasiepi, D. Lenz, K. Toledo, N., M. S. Bargo, and S. F. Vizcaíno. 2013. Tsangaras, B. Voirin, N. de Moraes-Barros, R. Muscular reconstruction and functional D. E. MacPhee, and A. D. Greenwood. 2016. morphology of the forelimb of early Miocene Evolutionary relationships among extinct and sloths (Xenarthra, Folivora) of Patagonia. extant sloths: The evidence of mitogenomes Anatomical record 296:305–325. and retroviruses. Genome Biology and Evolution 8:607–621. Toledo, P. M. 1996. Locomotor Patterns within the Pleistocene Sloths. PhD Thesis, University of Smith, F. A., S. K. Lyons, S. K. M. Ernest, K. E. Jones, Colorado. D. M. Kaufman, T. Dayan, P. A. Marquet, J. H. Brown, and J. P. Haskell. 2003. Body Mass of Tsegai, Z. J., T. L. Kivell, T. Gross, N. Huynh Nguyen, Late Quaternary Mammals. Ecology 84:3403. D. H. Pahr, J. B. Smaers, and M. M. Skinner. 2013. Trabecular bone structure correlates with Smith, P., and R. D. Owen. 2017. Calyptophractus hand posture and use in hominoids. PLoS ONE retusus (Cingulata: Dasypodidae). Mammalian 8. Species 49:57–62. Tsegai, Z. J., M. M. Skinner, A. H. Gee, D. H. Pahr, Sode, M., A. J. Burghardt, R. A. Nissenson, and S. G. M. Treece, J.-J. Hublin, and T. L. Kivell. Majumdar. 2008. Resolution dependence of the 2017. Trabecular and cortical bone structure of mon-metric trabecular structure indices. Bone the talus and distal tibia in Pan and Homo. 42:728–736. American Journal of Physical Anthropology 163:784–805. Stephens, N. B., T. L. Kivell, T. Gross, D. H. Pahr, R. A. Lazenby, J.-J. Hublin, I. Hershkovitz, and M. Tsegai, Z. J., M. M. Skinner, D. H. Pahr, J.-J. Hublin, M. Skinner. 2016. Trabecular architecture in the and T. L. Kivell. 2018. Systemic patterns of thumb of Pan and Homo: implications for trabecular bone across the human and investigating hand use, loading, and hand chimpanzee skeleton. Journal of Anatomy preference in the fossil record. American 232:641–656. Journal of Physical Anthropology 161:603–619. Vizcaíno, S. F., M. S. Bargo, and G. H. Cassini. 2006. Stock, C. 1942. A ground sloth in Alaska. Science Dental occlusal surface area in relation to body 95:552–553. mass, food habits and other biological features in fossil xenarthrans. Ameghiniana 43:11–26. Straehl, F. R., T. M. Scheyer, A. M. Forasiepi, R. D. E. MacPhee, and M. R. Sánchez-Villagra. 2013. Vizcaíno, S. F., R. A. Fariña, and G. V. Mazzetta. Evolutionary patterns of bone histology and 1999. Ulnar dimensions and fossoriality in bone compactness in xenarthran mammal long armadillos. Acta Theriologica 44:309–320. bones. PLoS ONE 8:e69275. Vizcaíno, S. F., N. Toledo, and M. S. Bargo. 2017. Su, A., and K. J. Carlson. 2017. Comparative analysis Advantages and limitations in the use of extant of trabecular bone structure and orientation in xenarthrans (Mammalia) as morphological South African hominin tali. Journal of Human models for paleobiological reconstruction. Evolution 106:1–18. Journal of Mammalian Evolution, doi: 10.1007/s10914-017-9400-2. Su, A., I. J. Wallace, and M. Nakatsukasa. 2013. Trabecular bone anisotropy and orientation in White, J. L. 1997. Locomotor Adaptations in Miocene an Early Pleistocene hominin talus from East Xenarthrans. Pp. 246–264 in R. F. Kay, R. H. Turkana, Kenya. Journal of Human Evolution Madden, R. L. Cifelli, and J. J. Flynn, eds. 64:667–677. Vertebrate paleontology in the neotropics. The Miocene fauna of La Venta, Colombia. Tacutu, R., C. T. Craig, A. Budovsky, D. Wuttke, G. Smithsonian Institution Press, Washington. Lehmann, D. Taranukha, J. Costa, V. E. Fraifeld, and J. De Magalhães. 2013. Human Woodburne, M. O. 2010. The Great American Biotic Ageing Genomic Resources: Integrated Interchange: dispersals, tectonics, climate, sea databases and tools for the biology and level and holding pens. Journal of Mammalian genetics of ageing. Nucleic Acids Research Evolution 17:245–264.

Amson & Nyakatura 2018 — Long bone structure in Xenarthrans 20